Non Uniformity Correction Using Cosine Functions and Total Variation Constraint in Musculoskeletal Nmr Imaging

Introduction: In this work, we attempted to correct the non-uniformities generated in lower limbs (calf and thigh) NMRI by RF inhomogeneities. Correction techniques are based in general on prior knowledge about the intensity distribution inside tissues (or tissue magnetic properties) which is not straightforward in case of pathological muscles, as in neuro-muscular disorders. For our application, we can rely only on the assumption of signal uniformity in the subcutaneous fat and we need to estimate the non uniformity values in muscles. An interpolation technique like spline or polynomial would not be efficient because the muscle area is large relative to fat. Parametric models are an attractive solution to the problem. Based on the observation that the non uniformity field is a low frequency signal, we expressed it as a finite sum of discrete cosine functions. The estimation of the parameters of the non uniformity field was done through the minimization of a convex cost function.